Poster anthonyevans asked over on Stockhouse whether one can infer apabetalone effectiveness in BETonMACE from just the gross data. I'm copying my answer here:

"This "simple" question has come up over and over again on various forums. Short answer is no.

Long answer. Too many unknown variables. BETonMACE modeled after patient population in the EXAMINE trial with the DPP4 inhibitor alogliptin (NEJM article) in which diabetic patients with recent ACS were tested. EXAMINE was run 2009-2013. The drug failed to reduce 3-point MACE. Both groups had 3-point MACE event rates around 7.5-7.9 events per 100 patient years. One unknown is that we can't assume that the baseline patient characteristics and placebo event rate will be the same between BETonMACE and EXAMINE. Additionally, BETonMACE has a low-HDL requirement that EXAMINE did not. So it could be that BETonMACE patients have an even higher baseline/placebo risk than patients in EXAMINE.

The only "official" update on event rate was first stated in the Feb 2018 BIO CEO slide 15: "Projected primary MACE rate still 8.0 per 100 patient years on top of aggressive standard of care = strong unmet need." Upon request for clarification, Clayton replied that "these numbers are derived from all patients" meaning that this is the blinded data that includes apabetalone and placebo. In June, Tundup posted that "The rate of MACE events has been dropping steadily since the beginning of the trial, starting off at the expected 8% per 100 patient years but now down to 7.2% per 100 patient years" but that came second hand from Tundup and not officially from a company communication. Either way, for a given event rate, there are multiple ways the placebo/apabetalone split could occur to get that same number. Let's go with 8% overall event rate for combined placebo and apabetalone groups. To get 8% overall event rate, it could be a 11% to 5% split, 10% to 6% split, 9% to 7% split, or no effect with an 8% to 8% split. 

"Can an inference be drawn from the Gross data?" No"

Breakdown of 3-point MACE: After posting this, it made me re-look at EXAMINE for the breakdown of 3-point MACE events. In EXAMINE, the median follow up period was 18 months. BETonMACE was originally going to be similar, but with the trial protocol amendment to allow dosing beyond 104 weeks the median follow up period will be a bit longer in BETonMACE. There were 621 3-point MACE events in EXAMINE (since no statistically significant effect of alogliptin, I am combining alogliptin and placebo groups here). Of the 621 3-point MACE events, 200 were CVD death, 360 were non-fatal myocardial infarction and 61 were non-fatal stroke. 

Low-HDL: In re-listening to the AGM webcast the other day, one comment from DM really caught my attention. In slide 14 in the BETonMACE trial design slide, DM states "The reason for the low-HDL is because it accounts for about 70% of patients with events." On clinical trials.gov, BETonMACE lists the low-HDL requirement as: "For males HDL-C<40 mg/dL(1.04 mmol/L), for females HDL-C<45 mg/dL(1.17 mmol/L) at Visit 1." Baseline HDL in EXAMINE was ~43 mg/dL (mean or average). Once the complete BETonMACE baseline patient data is presented at American Heart Association meeting next month, we can compare this to the baseline data from EXAMINE more closely. But the cognitive subroup baseline data and the CKD subgroup baseline data support that the BETonMACE inclusion criteria were met and that these BETonMACE patients are indeed very high risk. HDL levels in the cognition and CKD subgroups were 34 mg/dL (median). The mean HDL in BETonMACE should be lower than the mean 43 mg/dL in EXAMINE.

BearDownAZ